Fraud Detection in Mobile Payments Using a Hybrid Machine Learning Model

The rapid expansion of mobile payment systems has increased the risk of financial fraud, necessitating the development of advanced fraud detection mechanisms. This paper proposes a novel hybrid machine learning model that integrates supervised learning (XGBoost) and unsupervised anomaly detection (Isolation Forest) to enhance fraud detection accuracy while minimizing false positives. The model was trained and evaluated using a publicly available financial transaction dataset, achieving an accuracy of $99.1 \%$ and an AUC-ROC of 0.996, significantly outperforming traditional machine learning and deep learning models. The hybrid approach leverages behavioral and transactional attributes, ensuring adaptability to evolving fraud patterns. Extensive feature importance analysis highlights key fraud indicators such as failed login attempts, transaction amounts, and system latency, reinforcing the model’s reliability. Comparative performance analysis with state-of-the-art fraud detection techniques demonstrates the superior effectiveness of the proposed approach in identifying fraudulent transactions with minimal false alarms. The findings confirm that integrating ensemble learning techniques with anomaly detection provides a scalable and real-time fraud detection framework, suitable for deployment in financial institutions, fintech applications, and mobile banking platforms.